JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 107, NO. B12, 2381, doi:10.1029/2001JB000710, 2002

Crustal structure and relocated earthquakes in the Puget Lowland, Washington, from high-resolution seismic tomography T. M. Van Wagoner,1 R. S. Crosson, K. C. Creager, G. Medema, and L. Preston Department of Earth and Space Sciences, University of Washington, Seattle, Washington, USA

N. P. Symons Sandia National Laboratories, Albuquerque, New Mexico, USA

T. M. Brocher U.S. Geological Survey, Menlo Park, California, USA

Received 13 June 2001; revised 19 July 2002; accepted 21 August 2002; published 31 December 2002.

[1] The availability of regional earthquake data from the Pacific Northwest Seismograph Network (PNSN), together with active source data from the Seismic Hazards Investigation in Puget Sound (SHIPS) seismic experiments, has allowed us to construct a new high- resolution 3-D, P wave velocity model of the crust to a depth of about 30 km in the central Puget Lowland. In our method, earthquake hypocenters and velocity model are jointly coupled in a fully nonlinear tomographic inversion. Active source data constrain the upper 10–15 km of the model, and earthquakes constrain the deepest portion of the model. A number of sedimentary basins are imaged, including the previously unrecognized Muckleshoot basin, and the previously incompletely defined Possession and Sequim basins. Various features of the shallow crust are imaged in detail and their structural transitions to the mid and lower crust are revealed. These include the Tacoma basin and fault zone, the Seattle basin and fault zone, the Seattle and Port Ludlow velocity highs, the Port Townsend basin, the Kingston Arch, and the Crescent basement, which is arched beneath the Lowland from its surface exposure in the eastern Olympics. Strong lateral velocity gradients, consistent with the existence of previously inferred faults, are observed, bounding the southern Port Townsend basin, the western edge of the Seattle basin beneath Dabob Bay, and portions of the Port Ludlow velocity high and the Tacoma basin. Significant velocity gradients are not observed across the southern Whidbey Island fault, the Lofall fault, or along most of the inferred location of the Hood Canal fault. Using improved earthquake locations resulting from our inversion, we determined focal mechanisms for a number of the best recorded earthquakes in the data set, revealing a complex pattern of deformation dominated by general arc-parallel regional tectonic compression. Most earthquakes occur in the basement rocks inferred to be the lower Tertiary Crescent formation. The sedimentary basins and the eastern part of the Olympic subduction complex are largely devoid of earthquakes. Clear association of hypocenters and focal mechanisms with previously mapped or proposed faults is difficult; however, seismicity, structure, and focal mechanisms associated with the Seattle fault zone suggest a possible high-angle mode of deformation with the north side up. We suggest that this deformation may be driven by isostatic readjustment of the Seattle basin. INDEX TERMS: 7205 Seismology: Continental crust (1242); 7218 Seismology: Lithosphere and upper mantle; 7230 Seismology: Seismicity and seismotectonics; KEYWORDS: tomography, structure, Puget, earthquakes Citation: Van Wagoner, T. M., R. S. Crosson, K. C. Creager, G. Medema, L. Preston, N. P. Symons, and T. M. Brocher, Crustal structure and relocated earthquakes in the Puget Lowland, Washington, from high-resolution seismic tomography, J. Geophys. Res., 107(B12), 2381, doi:10.1029/2001JB000710, 2002.

1. Introduction 1Now at ChevronTexaco North America Upstream, New Orleans, [2] Although most damaging historical earthquakes in Louisiana, USA. the region have been within the Wadati-Benioff zone of Copyright 2002 by the American Geophysical Union. the Cascadia subduction zone (e.g., the 2001 magnitude 0148-0227/02/2001JB000710$09.00 6.8 Nisqually earthquake [Malone et al., 2001]), much

ESE 22 - 1 ESE 22 - 2 VAN WAGONER ET AL.: CRUSTAL STRUCTURE AND EARTHQUAKES

Figure 1. Schematic geologic map of northwestern Washington and southern British Columbia, showing the model region (heavy inner rectangle). Abbreviations for major cities: E—Everett; O— Olympia; S—Seattle; T—Tacoma; V—Victoria. Abbreviations for faults: CRBF—Coast Range boundary fault; DF—Doty fault; DDMF-Darrington—Devils Mountain fault; HCF—Hood Canal fault; HRF—Hurricane Ridge fault; LRF—Leech River fault; OF—Olympia fault; SCF—Straight Creek fault; SF—Seattle fault; SHZ—St. Helens zone; SJF—San Juan fault; SqF—Sequim fault; SWIF—Southern Whidbey Island fault; TF—Tacoma fault; WRF—White River fault. Abbreviations for oil test wells: DS—Dungeness Spit #1; MK—Mobil Kingston #1; P18—Pope-Talbot #18-1; SC—Silvana Community #12-1; SS—Socal Schroeder #1; SW—Socal Whidbey #1; WS—Washington State #1. Abbreviations for geologic, structural, and topographic features: BH—Blue Hills; BkH—Black Hills; DB—Dabob Bay; EB—Everett basin; HC—Hood Canal; KA—Kingston arch; LS—Lake Sammamish; LW—Lake Washington; M—Metchosin igneous complex; PTB—Port Townsend basin; PS—Puget Sound; SB— Seattle basin; SU—Seattle uplift; TB—Tacoma basin. Abbreviations for Cascade volcanoes: MSH—Mt. St. Helens; MR—Mt. Rainier; MB—Mt. Baker. Sources: Blakely et al. [2002]; Brocher et al. [2001]; Brocher and Ruebel [1998]; Gower et al. [1985]; Johnson et al. [1999]; Schruben et al. [1994]; Tabor and Cady [1978a]. recent geological and geophysical research in the Puget investigations [e.g., Brocher et al., 2001; Blakely et al., Lowland of western Washington State (Figure 1) has been 2002]. prompted by increasing concern about hazards from crustal [3] Lees and Crosson [1990] carried out the first 3-D earthquake sources. A postulated earthquake about 1100 tomography for the Puget Lowland region, but were limited years ago on the Seattle fault [Bucknam et al., 1992] and by relatively few earthquake observations and sparse station concern about other possible crustal faults [e.g., Johnson et spacing. Symons and Crosson [1997] developed new inver- al., 1996, 1999] in the Lowland region motivated the sion technology and applied this to an improved earthquake recent SHIPS active source experiments [Brocher et al., data set plus limited explosion data. In their study, shallow 1999, 2000], and analyses based on these and other resolution was again limited by lack of surface source to VAN WAGONER ET AL.: CRUSTAL STRUCTURE AND EARTHQUAKES ESE 22 - 3

Figure 2. Map of the Puget Lowland model area showing source/receiver locations used to construct the high resolution model. The map area is the heavy inner rectangle shown in Figure 1. Temporary Wet- SHIPS stations are shown as diamonds (à) and Wet-SHIPS air gun tracks are shown as thin line segments. Permanent PNSN seismograph stations are shown as black triangles. Station distributions for both the Dry-SHIPS and the 1991 experiment are shown as line segments and experimental sources are shown as white stars. Higher confidence 3-D relocated earthquake epicenters are shown as white circles (6), lower confidence events as white squares (5). Earthquake observations were recorded only by PNSN stations. surface receiver observations. Seismic reflection profiling our knowledge of the crustal structure in this region. In this has been interpreted for structural details in only the upper study, the simultaneous analysis of earthquake data and few kilometers at most of the crust [e.g., Johnson et al., active source observations has allowed us to develop a high 1994, 1996; Pratt et al., 1997]. More recently, a first arrival quality 3-D model to a depth of 30 km, considerably greater P wave tomography model for the upper 11 km of the crust than previously reported [e.g., Brocher et al., 2001], and to was constructed by Brocher et al. [2001] in the same region simultaneously relocate nearly 1000 regional earthquakes as our study using data from the initial SHIPS experiment. for tectonic interpretation. In these previous studies, the structural transition to mid- crustal depths has not been accurately imaged. Since the 1.1. Geologic Setting midcrust is likely to be the seat of important tectonic [5] The tectonics of the Pacific Northwest region are processes in this forearc region, potentially valuable struc- dominated by the oblique, northeastward subduction of tural information has been missing. the Juan de Fuca plate beneath western North America. [4] The availability of active source data from SHIPS as Subduction of oceanic lithosphere during the Tertiary has well as two earlier experiments [Miller et al., 1997; Parsons led to the formation of two major geologic provinces in the et al., 1999], together with high quality crustal earthquake region: the Coast Range province on the west, which data acquired for the last 20 years from the PNSN regional includes the Olympic subduction complex (OSC) plus the network (Figure 2), provides an ideal opportunity to refine Puget Lowland, and the Cascade province on the east ESE 22 - 4 VAN WAGONER ET AL.: CRUSTAL STRUCTURE AND EARTHQUAKES

(Figure 1) [Brandon and Calderwood, 1990; Snavely and fication of faults with possible reverse offsets [e.g., Johnson Wells, 1996]. et al., 1994, 1999; Brocher et al., 2001] is consistent with [6] In the OSC, metamorphosed Tertiary sedimentary the focal mechanisms of crustal earthquakes [Ma et al., rocks from the accretionary prism have been uplifted since 1996]. Northward compression is most likely the result of the Miocene to form the Olympic mountains [Tabor and oblique subduction [Wang, 1996], possibly in conjunction Cady, 1978a; Brandon et al., 1998]. Within the eastern with northward migration of the Cascadia forearc [Brandon Olympics is the steeply dipping Crescent Formation—thick and Calderwood, 1990; Wells et al., 1998]. submarine and subaerial Paleocene to Eocene basalts which [10] Geologic identification of shallow faults is hampered may have formed by extrusion associated with extension by thick glacial sediments covering much of the Lowland along the margin of north America [Wells et al., 1984; and diffuse crustal seismicity does not reveal distinct Babcock et al., 1992]. Olympic uplift has tilted the Crescent lineations that may be interpreted as active faults. Various to its present east-dipping position on the western boundary geophysical techniques, including gravity and magnetic of the Lowland [Tabor and Cady, 1978a, 1978b; Brandon surveys, seismic reflection profiles, and seismic tomogra- and Calderwood, 1990]. The Crescent formation is corre- phy, have been employed to identify faults and other lated with the Siletz River volcanics in Oregon [Snavely and structures [e.g., Danes et al., 1965; Blakely et al., 1999; Baldwin, 1948], the Metchosin igneous complex on south- Pratt et al., 1997; Brocher et al., 2001]. ern Vancouver Island [Massey, 1986], and with basalt out- [11] The most important shallow fault zone in the region crops in the Blue Hills east of Bremerton and in the Black is the E–W striking Seattle fault which runs directly Hills southwest of Olympia (Figure 1) [Babcock et al., beneath the Seattle metropolitan area and has been inter- 1994]. Thickness of the Siletz volcanics in Oregon may be preted to be a south-dipping thrust or reverse fault (Figure 1) as much as 34 km [Trehu et al., 1994], possibly thinning [Johnson et al., 1994, 1999; Pratt et al., 1997]. The Seattle northward into Washington where mapped thickness of the fault forms the southern boundary of the Seattle basin, an Crescent is more than 16 km [Babcock et al., 1992] and approximately 10 km deep basin composed of low velocity, tomographic imaging reveals Crescent extending to depths low density sedimentary rocks with a negative Bouguer as great as 25 km [Symons and Crosson, 1997]. gravity anomaly exceeding À120 mGal [Danes et al., 1965; [7] The eastern boundary of the Coast Range province is Finn et al., 1991; Johnson et al., 1994]. The upper 10–12 ill-defined beneath the Puget Lowland where it contacts the km of the crust within the Seattle fault zone have been pre-Tertiary basement of the Cascade province. The Cas- imaged by recent seismic reflection surveys and may consist cade basement comprises a variety of igneous, metamor- of multiple E–W oriented fault segments [Fisher et al., phic, and sedimentary rocks which were accreted to North 1999; Molzer et al., 1999; ten Brink et al., 2002]. South of America by the late Cretaceous or early Tertiary [Tabor, the Seattle fault, Crescent basement lies close to the surface 1994]. Cascade arc volcanism began in the late Eocene [e.g., Pratt et al., 1997]. Reverse displacement on the [Snavely and Wells, 1996]. Seattle fault has been interpreted to be responsible for both uplift of the basement south of the fault and subsidence of 1.2. Seismicity the Seattle basin north of the fault [Johnson et al., 1994, [8] Earthquakes in western Washington generally fall into 1999; Pratt et al., 1997]. two categories: crustal events which occur at depths less [12] Paleoseismic investigations revealed an uplifted than about 30 km, and subcrustal events associated with the marine terrace at Restoration Point on Bainbridge Island subducted oceanic lithosphere which typically occur at which suggests an earthquake, possibly of magnitude 7.0 or depths greater than 35 km [Ludwin et al.,1991].The greater, occurred on the Seattle fault approximately 1100 majority of crustal earthquakes in western Washington years ago [Bucknam et al., 1992]. Also within the Seattle occur beneath the Puget Lowland and the eastern edge of fault zone is a recently discovered north-dipping fault scarp the Cascade volcanic arc, with virtually none occurring in on southern Bainbridge Island with evidence of north-side- the OSC. The largest crustal earthquakes with modern up offset within the last 3–4 ka [Bucknam et al., 1999; network observations in the Puget Lowland are the magni- Nelson et al., 1999]. tude 5.0 Robinson Point event in 1995, the magnitude 5.4 [13] South of the Seattle fault zone, a bounding fault on Duvall event in 1996, and the magnitude 4.9 Bremerton the north side of the Tacoma basin was proposed by Danes earthquake in 1997. A probable crustal earthquake with an et al. [1965] from gravity data, and by Gower et al. [1985] estimated magnitude greater than 6.5 occurred in 1872 to based on gravity and aeromagnetic anomalies. Velocity the northeast of the study area in the North Cascades gradients from shallow tomography, in addition to docu- [Malone and Bor, 1979; Bakun et al., 2001]. Focal mech- mented uplift at two localities north of the Tacoma basin anisms for crustal earthquakes beneath the Puget Lowland [Bucknam et al., 1992; Sherrod, 1998], led Brocher et al. are mainly thrust and strike-slip, with shallowly plunging P- [2001] to interpret this boundary as a north-dipping reverse axes oriented north–south [Ludwin et al., 1991; Ma et al., fault, designated the Tacoma fault. 1996]. [14] Northward motion of the Cascadia forearc region during the early Tertiary may have been accommodated 1.3. Shallow Structure and Faulting along the proposed Coast Range Boundary fault (CRBF), a [9] The shallow structure of the Puget Lowland consists right-lateral strike-slip fault which presumably separates of sedimentary basins and relative basement highs rocks of the Coast Range terrane from the pre-Tertiary delineated mainly by gravity and seismic reflection obser- basement of the Cascades [Johnson, 1984, 1985; Johnson vations. Pratt et al. [1997] interpreted these structures to et al., 1996]. The CRBF is inferred mainly from sedimento- have formed in a compressive tectonic regime. The identi- logic evidence to lie beneath the eastern Puget Lowland VAN WAGONER ET AL.: CRUSTAL STRUCTURE AND EARTHQUAKES ESE 22 - 5 where it strikes approximately N–S (Figure 1). As a maximum RMS residual of 0.3 s, maximum estimated possible extension of the CRBF, the northern boundary errors in the hypocenter location of 2.5 km horizontally between the Coast Range basement and the Cascade block and 5.0 km vertically, maximum azimuthal gap of 135°, and is the southern Whidbey Island fault zone, a NW trending maximum distance to the nearest station of less than twice transpressional fault [Johnson et al., 1996], which may in the event depth (or 10 km, whichever is greater), or (3) more turn connect with the Leech River fault to the west on than 20 P-picks [Symons, 1998]. The high quality earth- southern Vancouver Island where the terrane boundary is quakes were declustered to remove redundancy and exposed at the surface (Figure 1) [Clowes et al., 1987; repicked to ensure accuracy, resulting in 555 events. The Johnson et al., 1996]. second group of earthquakes are medium quality and consist of 357 events with the following characteristics: (1) magni- tude greater than or equal to 2.0, and (2) at least 10 picks of 2. Data Description the P wave arrival, and (3) maximum azimuthal gap of [15] Data used to construct our model originated from 160°. Events in the medium quality database were also three active source seismic experiments, and from earth- repicked to ensure accuracy. quakes recorded by the permanent Pacific Northwest Seis- mograph Network (PNSN) operated by the University of Washington (Figure 2). 3. Tomography Method and Results [19] P wave first-arrival observations from both earth- 2.1. Active Source Data quakes and explosions were used to generate a P wave [16] Counting the number of source-receiver P wave velocity model of the Puget Lowland. The model is defined arrivals, eighty-one percent of the data originated from the within a rectangular box between longitudes 123.4°W and first phase of the Seismic Hazards Investigation of Puget 121.5°W, latitudes 46.9°Nand48.35°N, and extending Sound (SHIPS), nicknamed ‘‘Wet-SHIPS’’, during which from above the surface to 40 km depth (Figure 2). Model approximately 250 temporary land-based seismographs coordinates are given throughout the paper in kilometers were deployed to record more than 33,000 air gun sources east (E) and north (N) from 46.9°N and 123.4°W. For in the waterways of Puget Sound and the Straits of Juan de convenience, these are given in the text as coordinate pairs Fuca and Georgia (Figure 2) [Brocher et al., 1999]. Nearly in parentheses—for example: (80, 95), where the position is one million travel time picks from the Wet-SHIPS project 80 km east and 95 km north. Single coordinate values are were initially made by USGS and other project collabora- stated as kilometers east or north (e.g., 80 E or 40 N). The tors [Brocher et al., 2001]. In order to minimize redundancy velocity model is defined on a grid of nodes within the in the data and for computational efficiency, we divided our model space, with a node spacing of 1.5 km in both model area into 0.7 km2 grids and then selected one source horizontal and vertical dimensions, resulting in 331,360 in each grid element that had the largest number of total velocity model parameters. associated picks, resulting in a total of 64,675 first-arrival [20] Nonlinear 3-D seismic tomography was carried out picks which were used in our inversion. by the method of Symons and Crosson [1997], which jointly [17] The second phase of the SHIPS project was an E–W inverts arrival time observations from both earthquakes and land refraction experiment, nicknamed ‘‘Dry-SHIPS’’, dur- explosions to simultaneously obtain a P wave slowness ing which more than 1000 temporary land-based seismo- (inverse of velocity) model and relocated earthquake hypo- graphs were deployed across the central Puget Lowland to centers. The method is fully nonlinear, using linearized record 38 drill hole sources (Figure 2) [Brocher et al., iterations to achieve a minimum L2 norm between obser- 2000]. From the trace data we made 3588 first-arrival picks vations and theoretical predictions. Travel times are com- from 32 shots, all of which were included in the inversion, puted by finite difference using nodal parameterization. representing 4.5% of the total data. Also included in the Each constraint equation in the linearized system is inversion were 880 travel times from 5 land sources weighted by the inverse of its picking error, with regulari- recorded by 252 temporary land-based seismometers zation imposed to smooth the final model. We utilized a deployed during the 1991 N–S land refraction experiment discrete Laplacian regularization operator which uses the six (Figure 2) on the west flank of the Cascade Range [Miller et nearest neighbors on the rectangular grid of nodes. This al., 1997], and 39 travel times from three of these sources operator is anisotropic with a horizontal to vertical smooth- recorded by permanent PNSN stations. For all active source ing ratio of 5. The linearized system is solved at each data, water depth time corrections were applied and picking iteration by conjugate gradient least squares. The model and error is assumed to be 50 ms. hypocenters are updated after each iteration and the inver- sion is repeated until the norm of the model perturbation 2.2. Earthquake Data falls below a small predetermined value, typically in less [18] A total of 912 earthquakes, selected from the PNSN than 20 iterations. Since our model is fully three-dimen- catalog from 1980 through 2000, was used in the inversion sional, separate station corrections, which are commonly (Figure 2). Events prior to 1980 were not digitally recorded used to approximate lateral velocity variations near stations, and were not included in the inversion because of difficulty are not necessary and were not included in the model in verifying timing. Two earthquake databases were con- parameterization. structed with different selection criteria. The first group of [21] Initial earthquake hypocenters used in the inversion earthquakes are the highest quality and possess at least one were determined using the current PNSN Puget Sound of the following three characteristics: (1) magnitude greater velocity model and station corrections [Crosson, 1976; than or equal to 2.5, or (2) at least 10 P wave arrival picks, Ludwin et al., 1991]. A slightly modified version of the ESE 22 - 6 VAN WAGONER ET AL.: CRUSTAL STRUCTURE AND EARTHQUAKES

Table 1. 1-D Starting Model for model PS3DA is 97 ms—this represents a simple Depth Range, km P-Velocity, km/s residual variance reduction of 98%. 0.0–1.0 4.00 3.1. Resolution 1.0–5.0 5.80 5.0–10.0 6.60 [23] To determine the resolving ability of our model, we 10.0–15.0 6.75 performed a series of 3-D checkerboard tests in which 15.0–20.0 6.85 model PS3DA was perturbed by adding a 3-D sinusoidal 20.0–25.0 6.95 pattern with a maximum amplitude of 10% of the model 25.0–40.0 7.80 velocity value (maximum of ±0.81 km/s). The sinusoid wavelength is variable for different tests (Figure 3). Syn- thetic travel times were then generated through the per- PNSN 1-D model, based on preliminary inversion trials, was turbed model using the same source-receiver positions as used as our starting model for the 3-D inversion (Table 1). for the actual data. Using the smoothing parameters from [22] Several regularization weights were tried in order to our final model, we inverted the synthetic travel times, generate models with varying degrees of smoothness. Our allowing the hypocenters to be simultaneously relocated. preferred final model, designated PS3DA, had small overall The vertical sinusoid wavelength was fixed at 20 km, travel time residuals and is realistic compared to previous producing a vertical cell size of 10 km. Two separate models and to current knowledge of geologic structure inversions using both 20 km and 30 km horizontal wave- beneath the Lowland. Models with larger regularization lengths were performed, resulting in 10 and 15 km grids in weights had unreasonably large residuals and models with map view. smaller weights were unrealistically rough. The overall RMS [24] The checkerboard pattern has been recovered remark- residual for the best fitting 1-D model is 710 ms, while that ably well at both the 15 and 10 km cell spacing (Figure 3).

Figure 3. Recovered sinusoidal checkerboard perturbations from the final 3-D model using a vertical cell spacing of 10 km. (a) Horizontal slices with a lateral anomaly spacing of 15 km, (b) horizontal slices with a lateral anomaly spacing of 10 km, and (c) vertical sections through the 15 km lateral anomaly perturbation whose locations are shown by black pointers in (a). Grid lines show boundaries of original checkerboard pattern. Grid dimensions listed in black boxes are kilometers east, north, and depth respectively. VAN WAGONER ET AL.: CRUSTAL STRUCTURE AND EARTHQUAKES ESE 22 - 7

The model is best resolved laterally in the upper 10 km Table 2. Laboratory P Wave Velocities for Western WA ‘‘Mafic due largely to the active source constraint. At 5.5 km depth Rocks’’a the 15 km cell pattern and the perturbation amplitudes Pressure, Approximate Depth, Average P-Velocity, have been recovered well throughout the majority of the MPa km km/s model (Figures 3a and 3c). At this depth, cell spacing of 20 0.7 5.82 ± 0.51 10 km has also been recovered well with some loss of 40 1.5 5.93 ± 0.46 resolution on the western and southern edges and north 60 2.2 5.99 ± 0.42 and south of Lake Washington where ray coverage is 80 2.9 6.04 ± 0.41 100 3.6 6.07 ± 0.40 minimal (Figure 3b). 200 7.3 6.18 ± 0.36 [25] At 14.5 km depth, the 15 km cell spacing has been a Velocities are reported as averages and standard deviations of recovered throughout most of the model area and the measurements at each respective pressure [Brocher and Christensen, majority of the 10 km cell spacing in the central and eastern 2001]. parts of the model has also been recovered (Figures 3a and 3b). Observations at this depth and below originate mostly from earthquake sources, as most ray paths from the active wave velocities of 29 Crescent formation rocks collected source surveys do not penetrate to this depth. Smearing of from the Olympic Peninsula and the Blue Hills increase on velocities between adjacent cells at the 10 km spacing and no average from 5.8 to 6.2 km/s over confining pressures recovered checkerboard pattern beneath the western part of corresponding to depths of about 0.7–7.3 km (Table 2) the model area at 14.5 km depth reveals little to no resolution [Brocher and Christensen, 2001]. The standard deviations in these regions (Figure 3b). At 23.5 km depth, the 15 km of these velocities are about 10% of the means. These cell spacing in the central portion of the model has been values are consistent with published measurements from recovered where earthquakes are located (Figure 3a). global samples of unaltered basalts, while velocities for [26] Vertical anomalies of 10 km are recovered well zeolite to prehnite-pumpellyite facies basalt vary from 6.2 throughout the upper 30 km of the model (Figure 3c) where to 6.4 km/s over the same temperature and pressure range the recovered checkerboard pattern reveals a concave- [Christensen and Mooney, 1995]. Although Crescent upward zone which closely parallels the distribution of basalts contain zeolite and prehnite-pumpellyite alteration earthquake hypocenters. Perturbations in the upper 2–5 minerals, this formation also includes basaltic conglomerate km of the model are not fully recovered due to ray coverage and interbedded marine sedimentary units [Tabor and that is primarily focused around stations. Two anomalous, Cady, 1978a] which reduce its expected P wave velocity. high velocity regions are observed in E–W cross-section Based on these laboratory and borehole observations, we (Figure 3c) in the upper few kilometers due to overshoot expect the Crescent formation to have velocities of 4.5–5.5 effects near the model edges where observational con- km/s at depths less than 5 km, and velocities of 5.5–6.5 straints are lacking. The model is poorly resolved deeper km/s at depths greater than 5 km. Sonic speeds of sedi- than 30 km due to the sparse occurrence of earthquakes mentary and some volcaniclastic rocks in these exploration below that depth. wells typically range from 1.5 to 4.5 km/s; we thus interpret model velocities less than 4.5 km/s to represent sedimen- 3.2. Comparisons tary rocks. 3.2.1. Borehole Sonic Data and Laboratory Velocities 3.2.2. Refraction Profiles [27] Since direct verification of tomography models is [30] The profile analyzed by Schultz and Crosson [1996] difficult, we compared model PS3DA with measured sonic slants across the southern portion of our model and is velocity logs from six oil exploration wells in the Puget roughly comparable to Figure 4a. Although the resolution Lowland whose locations are shown in Figure 1 [Brocher of this profile cannot approach the 3-D tomography reso- and Ruebel, 1998]. Based on this comparison, model lution, velocities at 10–15 km depth are comparable at PS3DA velocities closely match the mean measured sonic approximately 6.5 km/s indicating good average agreement. speeds within the upper 2–3 km of the crust for the few The N–S profile along the Cascade range front analyzed by sites in the central portion of the model. Miller et al. [1997] lies about 15 km east of the section [28] The top of the Crescent formation is encountered in shown in our Figure 5c. Velocity gradients and values are in the Dungeness Spit, the Mobil Kingston #1, and the Pope generally good agreement between this profile and our Talbot #18-1 wells at approximately 1.5, 2.2 and 1 km model in the upper 20 km of the crust; however, the depth, respectively. Sonic speeds increase by 50–60% to Muckleshoot basin, which is readily apparent on our 4.0–6.0 km/s where the basalt is reached. At the depths of cross-section, is not apparent on the Miller et al. [1997] this contact, PS3DA velocities are averages of the shallow profile, probably because their profile lies almost entirely low velocity rocks and the higher velocity basalt, yielding east of the basin. velocities in the range of 3.5–5.0 km/s. Measured sonic speeds in the Kingston well are unavailable at the depth of the contact, but our model velocities are between 3.5 and 4. Interpretation and Discussion 4.0 km/s. Although these wave speeds slightly underesti- 4.1. Velocity Model mate the likely velocity of the basalt, model PS3DA [31] Our final P wave velocity model (PS3DA) is illus- velocities increase sharply below this depth to exceed 4.5 trated in a series of color maps and cross-sections (Figures km/s at 3.5 km depth. 4–6, including computed earthquake locations, with color [29] In general, the Crescent basement is deeper than 2 scales to represent velocity intervals. In this subsection we km and we expect its velocity to exceed 4.5 km/s. Mean P describe important features of the velocity model and our ESE 22 - 8 VAN WAGONER ET AL.: CRUSTAL STRUCTURE AND EARTHQUAKES

Figure 4. East–West cross-sections through the velocity model showing 3-D earthquake hypocenters within ±5 km of respective vertical section. Coordinates of cross-section locations are noted in lower right corner and shown in Figure 6 as red pointers. Structural abbreviations: CVH—Carnation velocity high; Cr—Crescent Fm (at surface); F—unnamed fault. Remaining structural abbreviations defined in Figure 5. Earthquake magnitude scale shown in the lower left corner. VAN WAGONER ET AL.: CRUSTAL STRUCTURE AND EARTHQUAKES ESE 22 - 9

Figure 5. North–South cross-sections through the velocity model showing 3-D earthquake hypocenters within ±5 km of each respective vertical section. Coordinates of cross-section locations are noted in lower right corner and shown in Figure 6 as red pointers. Structural abbreviations: KA—Kingston arch; MB— Muckleshoot basin; PB—Possession basin; PLVH—Port Ludlow velocity high; PTB—Port Townsend basin; SB—Seattle basin; SFZ—Seattle fault zone; SqF—Sequim fault; SVH—Seattle velocity high; SWIF—Southern Whidbey Island fault zone; TFZ—Tacoma fault zone; TB—Tacoma basin. Earthquake magnitude scale shown in the lower left corner. ESE 22 - 10 VAN WAGONER ET AL.: CRUSTAL STRUCTURE AND EARTHQUAKES

Figure 6. Horizontal depth slices through the 3-D model showing P wave velocities and 3-D earthquake hypocenters within ±1 km of each respective horizontal section. Model coordinates are measured in kilometers North and East from 123.4°W longitude and 46.9°N latitude. Velocities in regions of the model unconstrained by observations are not plotted. Red pointers show locations of cross-sections shown in Figures 4 and 5. Earthquake magnitude scale is shown in the lower left corner. Note the different velocity scale for e–h. VAN WAGONER ET AL.: CRUSTAL STRUCTURE AND EARTHQUAKES ESE 22 - 11

Figure 6. (continued) interpretation of significant structure as summarized in which is multilobed and elongated NW [Finn et al., 1991]. Figure 7. The Tacoma basin can be divided into two regions—one 4.1.1. Tacoma Basin and Fault portion approximately 25 by 35 km underlying south Puget [32] The Tacoma basin, near model coordinates (50, 40), Sound, and a NW trending linear segment approximately 15 is characterized by a negative Bouguer gravity anomaly by 30 km underlying the southern portion of Hood Canal ESE 22 - 12 VAN WAGONER ET AL.: CRUSTAL STRUCTURE AND EARTHQUAKES VAN WAGONER ET AL.: CRUSTAL STRUCTURE AND EARTHQUAKES ESE 22 - 13

(Figures 6a, 6b, and 6c). In N–S cross-section, the main 4 km depth (Figures 6c and 6d), which we interpret to be segment of the basin is asymmetric, with low velocity rocks the Crescent formation due to its similarly high velocity deeper to the north (Figure 5a). The thickness of sedimen- (5–6.5 km/s) and its traceable surface exposure west of tary units in the Tacoma basin was estimated to be 3.5 km Hood Canal. This relationship suggests direct structural by Pratt et al. [1997] based on seismic reflection profiles. connection between uplifted Crescent formation on the The basin reaches depths of 5–7 km (5.5 km/s velocity Olympic Peninsula and the Crescent rocks of the SVH. contour, Figures 4a and 5a) in our model and is truncated to 4.1.3. Muckleshoot Basin the north by steep velocity gradients (Figure 5a, 55 N) [35] Southeast of the SVH, near model coordinates (95, which we interpret to be a fault due to its coincidence with 45), is a previously undescribed basin approximately 20 by sharp gravity and magnetic anomalies [Danes et al., 1965; 25 km in size which corresponds to a localized Bouguer Finn et al., 1991; Blakely et al., 1999]. It was identified as a gravity anomaly in excess of 95 mG [Finn et al., 1991]. fault by Danes et al. [1965] and designated the Tacoma We informally designate this as the Muckleshoot basin fault by Brocher et al. [2001]. Vertical offset on the (Figures 6a–6d), after the Muckleshoot Native American Tacoma fault appears to be several kilometers, but the dip Reservation which is located near the center of the basin. of the fault cannot be confidently resolved by tomography. The Washington State borehole penetrates nearly 4 km of We have also imaged the eastern boundary of the north- Eocene and younger sedimentary and volcaniclastic rocks in trending, narrow-lobed portion of the basin (Figures 6a–6c, the southern end of the Muckleshoot basin (Figure 1) coordinates (30, 60)). The sharp velocity gradient along this [Johnson et al., 1994; Brocher and Ruebel, 1998]. Although boundary may represent a northward continuation of the ray path constraint is poor in the top few kilometers, and Tacoma fault as suggested by Brocher et al. [2001]. resolution in the vicinity of the basin is reduced, particularly 4.1.2. Seattle Velocity High (SVH) at the 10 km cell spacing (Figure 3b), the independent [33] The SVH is a NW trending block of high velocity observations of gravity and borehole data are consistent rock at relatively shallow depth (1–3 km) that separates the with the shape and depth of this low velocity feature from Seattle and Tacoma basins (Figures 6a–6c, coordinates (60, our tomography. 60)). This structural unit is called the ‘‘Seattle uplift’’ by [36] Above 4 km depth, velocities in the Muckleshoot several authors [e.g., Pratt et al., 1997] and interpreted as basin are higher than rocks at similar depths in the adjacent elevated Crescent basement with exposures of Crescent on Seattle and Tacoma basins. However, at 7 km depth the the west side of the unit in the Blue Hills southwest of Muckleshoot basin velocity is lower than that of the Tacoma Bremerton (Figure 1) [Yount and Gower, 1991; Haeussler basin. The eastern boundary trends N–S beneath the foothills and Clark, 2000]. Since there is no direct evidence of major of the Cascades where higher velocity rocks (5–6 km/s), basement uplift in the SVH, we prefer the tectonically neutral most likely representing the Cascade basement, are encoun- name of velocity high rather than uplift for this feature. The tered. From model PS3DA, the basin reaches a maximum maximum surface elevation of the SVH is 537 m at Gold depth of approximately 7–9 km (Figure 4a), approximately Mountain and for most of the structural block, basement 80% of the depth of the Seattle basin. velocity is reached at 1–3 km depth. The Bouguer gravity 4.1.4. Seattle Fault Zone (SFZ) over the SVH is close to zero [Danes et al., 1965; Finn et al., [37] The northern boundary of the SVH is marked by 1991] suggesting that this region, which is topographically steep velocity and gravity gradients corresponding to the near sea level, is close to isostatic equilibrium. SFZ. Seismic reflection profiles in Puget Sound reveal a 5- [34] The tomographic image of the SVH is a 50 by 25 km km-wide deformation zone within the upper few kilometers block with what appears to be a thin (1–2 km) sediment of the SFZ, interpreted by Johnson et al. [1994, 1999] and package overlying Crescent basement on the eastern portion, by Pratt et al. [1997] as a multisegmented, southward- and limited surface exposure of Crescent on the western dipping thrust fault. Our tomography results suggest that the portion (Figures 5a and 5b). The southern boundary of the main SFZ strikes WNW (Figures 6b and 6c) and is SVH is the proposed Tacoma fault, which separates high subparallel to the Tacoma fault. In cross-section, velocity velocity basement rock from the lower velocity rocks of the contours are nearly vertical in places where structural relief Tacoma basin. The western portion of the SVH contains two is the greatest (Figure 5b, 70 N). Although resolution from subcircular high velocity features (Figure 6c, near coordi- tomography is not sufficient to determine fault dip, our nates (45, 60)), coincident with magnetic highs [Blakely et results are consistent with a high-angle boundary, and al., 1999], the northern-most of which reaches the surface in possibly a thrust fault with southward dip. The steepest the Blue Hills. The northwest corner of the SVH appears to lateral velocity gradient is visible in N–S cross-section be continuous with rocks west of Hood Canal at and below beneath Puget Sound from Bremerton to West Seattle

Figure 7. (opposite) Interpretive horizontal section showing the velocity model at 2.5 km depth. Thick dashed red lines show our interpretation of fault locations from the tomography; thin dashed red lines show locations or limits of faults inferred from Johnson et al. [1994, 1996, 1999] and Gower et al. [1985]. The Toe Jam Hill fault scarp is labeled and shown as a solid red line. Blue triangles pointing up (Á) are locations of uplift and blue triangles pointing down (5) are locations of subsidence dated approximately 1100 years ago [Bucknam et al., 1992]. All relocated earthquakes are shown as open circles with magnitude scale in the lower left corner. Abbreviations: BH—Blue Hills; CVH—Carnation velocity high; DB—Dabob Bay; DDMF—Darrington-Devils Mountain fault zone; OF—Olympia fault; PB—Possession basin; PLVH— Port Ludlow Velocity High; PTB—Port Townsend basin; RP—Restoration Point; SqB—Sequim basin; SqF—Sequim fault; SVH—Seattle velocity high; SWIF—Southern Whidbey Island fault zone. ESE 22 - 14 VAN WAGONER ET AL.: CRUSTAL STRUCTURE AND EARTHQUAKES

(62–75 E) and persists to a depth of approximately 10 km. [40] The eastern boundary of the PLVH is a steep velocity Velocity gradients across the fault west of 62 E are more gradient which trends NNW and was interpreted as a gradual except directly north of the Blue Hills (45 E) where possible fault by Brocher et al. [2001]. The steepness of contours are nearly vertical for 2–3 km. We are unable to this gradient most likely represents a fault (Figures 4c and identify vertical displacement on the SFZ any further west 7) that does not appear to extend further south than the than Hood Canal near coordinates (40, 80). At this location, PLVH or further north than the Sequim fault. Crescent a strong lateral velocity gradient forms the abrupt western basalts are exposed at the surface in a quarry in the south- boundary of the Seattle basin which we interpret to be the east corner of the PLVH. Basalts were also encountered at western end of the SFZ. Velocity contours east of Lake 1 km depth in the Pope-Talbot #18-1 well, and only a few Washington reveal limited vertical relief across the fault kilometers to the southeast on the Kingston arch at the (Figure 5c) and there is little evidence for structural relief Mobil-Kingston #1 well, the basalts are more than 2 km east of 100 E, which we interpret to be the eastern extent of deep (Figure 1) [Brocher and Ruebel, 1998]. These obser- the tomographically resolved SFZ. vations provide an independent constraint for the varying 4.1.5. Seattle Basin depth and composition of this basement structure. Due to a [38] The Seattle basin contains Eocene–Oligocene and lack of ray path constraint north of Dabob Bay, the younger sedimentary rocks, and has a negative Bouguer boundary between the Crescent basalts exposed at the gravity anomaly exceeding 120 mG [Danes et al., 1965; surface on the Olympic Peninsula and the western portion Finn et al., 1991]. Based on high-resolution seismic reflec- of the PLVH (Figures 4c, 6b, and 6c) cannot be resolved in tion profiling from the SHIPS experiment plus gravity the upper 2–3 km. However, below 2 km depth velocities modeling, ten Brink et al. [2002] detailed the evolution of north of Dabob Bay are slightly less than adjacent velocities the Seattle basin and fault, with deformation beginning in in the PLVH to the west. At 4 km depth and below, rocks in Middle Miocene. In the upper 2 to 3 km of model PS3DA, the PLVH are continuous with Crescent rocks northwest of the dimensions of the Seattle basin are approximately 70 km Dabob Bay. E–W and 25 km N–S, extending from Hood Canal to [41] The southern boundary of the PLVH was interpreted nearly the Cascade foothills (Figures 6a and 6b). The to be a fault by Brocher et al. [2001] (the Lofall fault). The deepest part of the Seattle basin, as defined by the 5.5 velocity gradients observed at this location are less than km/s P wave iso-velocity surface, occurs beneath Seattle at those areas in our model where faults are more likely to 9–10 km depth (Figure 6e, coordinates (80, 70)). Deeper exist. As a result, we cannot distinguish from the model than 7 km, the basin has a NE trending arcuate structure, alone whether this feature is a fault or a fold. The southeast which may reflect the extension of the basin around the east corner of the PLVH merges with the Kingston arch, an E– end of the Kingston arch (Figures 6d and 6e). Previous W trending shallow high velocity anomaly which bounds estimates of the basin-depth have been 7.5 km [Pratt et al., the northern Seattle basin and is interpreted as an anticlinal 1997] and 9–10 km [Danes et al., 1965; Johnson et al., fold [Johnson et al., 1994, 1996; Pratt et al., 1997]. We 1994; Brocher et al., 2001]. The northern boundary of the have imaged the Kingston arch as a WNW trending feature Seattle basin in our model is formed by the Port Ludlow approximately 10 by 25 km in dimension (Figures 6a–6c velocity high and the Kingston arch [Pratt et al., 1997; and 7), which appears to plunge eastward, consistent with Brocher et al., 2001]. The western edge of the basin appears the interpretation of Gower et al. [1985], and appears to be to be sharply truncated beneath Dabob Bay, coincident with the eastward extension of the PLVH. Constant velocity a sharp NNW trending velocity gradient (Figures 6a–6c) surfaces dip gently southward (10°–20°) on the south side which we interpret to be a fault. of the arch, and higher velocity Crescent rocks are at 2.1 km 4.1.6. Port Ludlow Velocity High (PLVH), Kingston depth at the crest as identified in the Mobil-Kingston #1 Arch, and Possession Basin borehole (Figure 1). The eastward-dipping structure of the [39] The PLVH is centered at coordinates (50, 115) in the arch is distinct from the basin velocities to approximately northwestern portion of model PS3DA (Figures 6a–6c), and 80 E at 6 km depth (Figure 6c). corresponds to a roughly rectangular, and slightly positive, [42] There are several similarities between the PLVH and Bouguer gravity anomaly [Finn et al., 1991]. Velocities the SVH to the south. Both structures are roughly rectan- within the PLVH are as high as 6.0 km/s at 4 km depth gular and have similarly high velocities representing rela- which we interpret as Crescent basement at shallow depth. tively shallow Crescent basement with sedimentary basins This feature was termed the ‘‘Port Ludlow uplift’’ by bounding three sides. Both features appear to be connected Brocher et al. [2001], but, like the SVH, we prefer the westward to the uplifted Crescent formation on the Olympic descriptive name PLVH. The northern boundary of the Peninsula. PLVH corresponds to the Sequim fault, an E–W trending [43] The northern slope of the Kingston arch forms the structure described in part by Johnson et al. [1996] and southern boundary of a small unnamed basin composed of named by Brocher et al. [2001]. Using the 4.5 km/s velocity Eocene–Oligocene to Quaternary sedimentary rocks [John- contour as an approximate guide to the location of the son et al., 1996; Pratt et al., 1997] which we informally Sequim fault, we trace it to the west from the PLVH as an designate as the Possession basin (coordinate (70, 110)). arcuate feature lying along a sharp velocity gradient with This basin is slightly elongated E–W and is approximately relative north side down displacement. The fault appears to 25 by 15 km (Figures 4c, 5b, 6a–6c, and 7). The western bound the south end of the Port Townsend basin and can be boundary of the basin is defined by distinct gravity and traced no further west than 20 E. This location for the fault velocity gradients where the basin is structurally juxtaposed is consistent with outcrops of basalt to the south near to the PLVH. The northern boundary of the basin is not well Discovery Bay [Tabor and Cady, 1978a]. imaged in the upper 1–2 km of our model, but seismic VAN WAGONER ET AL.: CRUSTAL STRUCTURE AND EARTHQUAKES ESE 22 - 15 reflection data suggests that the N and NE boundaries of the Velocity magnitudes may be overestimated by a few percent basin are formed by the southern Whidbey Island fault but velocities of 7 km/s and above are too high to represent [Johnson et al., 1996]. The basin does not appear to extend basalt. It is also unlikely that we have imaged the base of the much further east than Puget Sound (Figure 6), but ray path continental crust at these depths, which is more than 40 km coverage is poor in this part of the model. We estimate the deep on the west flank of the Cascades where upper mantle maximum basin depth to be approximately 5–6 km at velocities are 7.6–7.8 km/s or higher [Schultz and Crosson, coordinates (70, 112) (Figure 5b). 1996; Miller et al., 1997]. Gabbroic complexes have been 4.1.7. Southern Whidbey Island and Coast Range observed in the Blue Hills within the SVH and within the Boundary Faults ophiolite stratigraphy of the Metchosin igneous complex on [44] The southern Whidbey Island fault zone (SWIF), Vancouver Island, both of which have been correlated with proposed by Johnson et al. [1996] from seismic reflection the Crescent formation [Massey, 1986; Haeussler and and gravity data and correlation of boreholes (Figure 1), is Clark, 2000]. In addition, P wave velocities of gabbro from interpreted to represent the crustal boundary between the 5 to 35 km depth range from 6.8 to 7.2 km/s [Christensen pre-Tertiary Cascade basement on the northeast and the and Mooney, 1995], consistent with observed high veloc- Coast Range basement on the southwest. The Cascade ities in model PS3DA. We therefore interpret velocities in basement was penetrated northeast of the SWIF in the excess of 7 km/s in our model as higher velocity gabbro Silvana Community #12-1 borehole (Figure 1) at 2.2 km emplaced within the Crescent formation, an interpretation depth and the measured P wave velocity was 5.5 km/s on also made by Parsons et al. [1999]. average [Brocher and Ruebel, 1998]. Unfortunately, the 4.1.9. Additional Structures similar velocities of Crescent and Cascade basement rocks [48] The existence and inferred location of the Hood does not provide enough velocity contrast to clearly dis- Canal fault is based on gravity and topographic features tinguish between these two provinces using tomography. along Hood Canal (Figure 1) [Gower et al., 1985]. This The location of the SWIF, described by Johnson et al. fault presumably separates the Crescent formation exposed [1996], appears in our model to form the northeastern at the surface west of Hood Canal from Quaternary sedi- boundary of the Port Townsend basin (Figures 6a–6c, 7) mentary rocks to the east. E–W cross-sections through the where it trends NW toward Victoria (Figure 1) [Brocher et model reveal the Crescent basement dipping uniformly al., 2001]. In cross-section, however, this boundary appears eastward from surface exposures west of Hood Canal to be gradual (Figure 5a). Tracing the inferred location of (Figure 4b, 30 E) with no apparent vertical offset, raising the fault to the southwest in model PS3DA, no obvious questions about the existence of the Hood Canal fault. vertical structure on the fault is apparent. Although strike-slip or small vertical offsets at this location [45] The existence and general location of the Coast cannot be ruled out by our tomography model, the inferred Range boundary fault (CRBF) has been inferred by John- east-side-down structure along the full length of Hood son [1984, 1985] to lie beneath the central Puget Lowland Canal [Brocher et al., 2001] is not supported by our model. (Figure 1) based on tectonic and sedimentologic evidence. However, along a 10–15 km length stretch on the west side This fault is hypothesized to represent the contact between of Dabob Bay (coordinates (40, 95)) a steep velocity Crescent basement to the west and pre-Tertiary Cascade gradient is consistent with faulting in which the east side basement to the east. Velocity contrasts across the contact is displaced downward with respect to near surface Crescent may be minimal because of similar basement velocities formation on the west side (Figures 6a and 6b). This fault and comparable velocities in the shallow structure, partic- segment may be the west margin of the Seattle basin and it ularly if the fault has been inactive since the Eocene may be effectively a short continuation of the Seattle fault [Johnson et al., 1996]. Our model, PS3DA, shows no northwestward from its western end. North of Dabob Bay, it basement velocity contrast at the location of the hypothe- is unclear from the tomographic model whether this struc- sized CRBF. ture continues northward into the inferred location of the 4.1.8. Crescent Basement Hood Canal–Discovery Bay fault zone [Johnson et al., [46] Velocities in the shallowest regions of the model 1996], but our model does not suggest a significant lateral west of Hood Canal and in the Blue Hills range from 5.0 to velocity gradient in this region. 6.0 km/s where the Crescent formation is exposed at the [49] Northwest of the PLVH and west of the Sequim fault surface (Figure 1), consistent with our interpretation of is a small, low velocity structure which we informally Crescent basalt velocities in the upper 5 km of the model. designate as the Sequim basin (Figures 6a and 6b, coor- At 4 km depth (Figure 6c) we interpret velocities in excess dinates (13, 35)). This basin was also imaged by Rama- of 5.0 km/s in the western half of the model to be Crescent chandran [2001] in a study utilizing SHIPS data. The basin basement, which are most prominent west of Hood Canal, is elongated WNW and corresponds to a similarly shaped and in the SVH and PLVH. In E–W cross-section (Figure Bouguer gravity anomaly [Finn et al., 1991]. We are unable 4b), the Crescent basement forms a concave-upward struc- to estimate the lateral dimensions of the basin due to ture which can be traced to outcropping basement in the insufficient ray path coverage on the south, but the gravity eastern Olympics. observations suggest that this feature may extend westward [47] The Crescent basement appears to extend to the beyond the edge of our model. The northern boundary of maximum depth resolved in our model, giving it a thickness the basin is a fairly sharp velocity and gravity gradient of 20 km or more over much of our model area (Figures 4 which may be an E–W trending fault, with south-side- and 5). Below 7 km depth, the model is dominated by down. Near-surface velocities on the north side of this velocities in excess of 5.5 km/s, with localized anomalies gradient are in excess of 4.5 km/s, and we interpret them having velocities as high as 7.0–7.5 km/s (Figures 6d–6h). to be Crescent basement, consistent with the nearby Dung- ESE 22 - 16 VAN WAGONER ET AL.: CRUSTAL STRUCTURE AND EARTHQUAKES eness Spit borehole (Figure 1) where basalt was encoun- borders of the model on the north and east, mostly located tered at 1.5 km depth. We estimate the basin to have a within the Cascade terrane. maximum depth of 3 to 4 km (Figure 6). 4.2.2. Focal Mechanism Determination [50] Lack of ray coverage in the upper 5–6 km of the [54] A subset of relocated earthquakes were selected for northeastern region of our model prevented us from imaging focal mechanism analysis which were either: (1) included much of the Everett basin. The shallowest structure of the within the Seattle, Tacoma, or southern Whidbey Island basin is visible in Figure 6a where it appears as a NW fault zones with magnitudes greater than 2.2, or (2) had trending elliptical feature, northeast of the SWIF (Figure 7). magnitudes greater than 3.0, regardless of location. The The limited tomography picture is consistent with Bouguer three fault zone regions are illustrated by dashed rectangles gravity observations [Finn et al., 1991]. on Figure 8 and the earthquakes greater than magnitude 3 [51] Approximately 20 km east of the Seattle basin at the are listed in Table 3. Focal mechanisms for these events foothills of the Cascades, near coordinates (120, 90), is a were determined by P wave raytracing using one of two 1-D subcircular high velocity anomaly approximately 10 km in velocity models, chosen to be representative of the 3-D diameter. We will refer to this anomaly as the Carnation velocity structure depending on the location of the event. velocity high (CVH, Figures 4b, 6a, 6b, and 7). Although it The model used for events in the Seattle and Tacoma fault corresponds well with the general increase in velocity in this zones was derived by fitting linear velocity model trends part of the model, the CVH includes an apparently isolated through vertical samples taken from six central locations in core, extending 5 to 7 km in-depth, with velocities as high the 3-D model (Figure 8; ‘‘Seattle Model’’ in Table 4). as 7.0 km/s. Our resolution tests suggest good resolution in Mechanisms for events outside the Seattle and Tacoma fault this region at the 15 km grid spacing (Figure 3a) with zones were determined using samples taken from four somewhat reduced resolution at 10 km grid spacing. Further outlying and two central locations in the model (‘‘North/ investigation of the data constraints on this anomaly have East Model’’ in Table 4). failed to reveal any potentially corrupted data that might 4.2.3. Regional Seismicity and Focal Mechanism explain the existence of this feature. Hence, we tentatively Characteristics accept this anomaly as real. The CVH does not correspond [55] Earthquake locations are shown with magnitude to any known gravity or magnetic anomalies that could coding on Figures 6–7 and Figure 9. The most concentrated confirm its existence, and there are no mapped mafic seismicity occurs between 15 and 25 km depth within what exposures in its immediate vicinity. However, small out- we interpret to be the Crescent basement in the central part crops of metagabbro are mapped approximately 20 km to of the model (Figure 9b). Earthquakes are generally sparse the east and southeast of this feature [Tabor et al., 1993] and in the upper 10 km of the crust, but in the central part of the such rocks could produce these high velocities if they were model, corresponding to the SVH where Crescent basement more extensive in the subsurface. approaches the surface, hypocenters occur at shallower depths. In the central and western parts of the model, at 4.2. Relocated Earthquakes and Focal Mechanisms depths less than 10 km earthquakes are nearly absent north 4.2.1. Earthquake Locations of the SFZ, and virtually no seismicity is observed south of [52] A total of 912 earthquakes were relocated in the joint the Tacoma basin (Figures 9a and 9b). In the westernmost tomographic inversion. Comparison of the 3-D locations region (Figure 9a), the base of crustal seismicity decreases with the starting 1-D locations reveals a mean absolute in-depth northward to approximately 130 N. In the central epicentral change of 970 m and a deepening of hypocentral part of the model, shown in Figure 9b, there is a distinct depths of 235 m on average. Since the average distribution concave upward pattern in the deepest crustal seismicity of hypocenters did not change significantly from 1-D to 3-D south of the SWIF, with the largest concentration of events models, the 1-D model appears to be a good regional beneath the Seattle basin. Pratt et al. [1997] proposed a average of the 3-D model. Our earthquake data set contains midcrustal decollement as a fundamental component of their only well located, typically larger and tectonically signifi- thrust sheet hypothesis for the origin of the Puget Lowland cant events selected from the complete catalog of PNSN faults and basins. The seismicity pattern and velocity earthquakes. Comparison of our resulting location distribu- structure shows no direct evidence for this midcrustal tion with the complete catalog distribution containing feature, so if decoupling occurs on a subhorizontal decolle- smaller and less well located earthquakes indicates that ment, it has no apparent effect on the contemporary earth- the events in our data set are representative of the overall quake activity and is without clear structural manifestation. pattern of seismicity for the model region. [56] East of Lake Washington (Figure 9c), the zone of [53] Since our model region is smaller than the PNSN concentrated seismicity observed in the Crescent basement regional network in western Washington, some network is absent and earthquakes are more uniformly distributed stations normally used for 1-D model locations were from near the surface to more than 25 km depth. We excluded from the inversion. As a result, the distance to interpret this change to reflect the eastward transition into the nearest station and the azimuthal gap for some earth- the Cascade basement, which is compositionally distinct quakes was degraded. We recalculated the azimuthal gaps from the Coast Range basement and in most places lacks a for the 3-D hypocenters and determined that 179 events thick sedimentary cover. A similar change in seismicity is (approximately 20%) had gaps larger than 160°. Our con- observed north of the SWIF where the number of earth- fidence in the locations of these events is somewhat reduced quakes decreases and the events are more uniformly dis- and they are distinguished on Figure 2 from events with tributed in depth (Figures 9a and 9b). We believe these gaps less than 160°, which we classify as well located. changes may also be a seismic signature of the northward Nearly all of these lower confidence events are near the transition to the Cascade province. Shallow earthquakes VAN WAGONER ET AL.: CRUSTAL STRUCTURE AND EARTHQUAKES ESE 22 - 17

Figure 8. Map of selected earthquakes with well constrained focal mechanisms plotted using equal- area, lower-hemisphere projections. Numbered mechanisms are events greater than magnitude 3 corresponding to events listed in Table 3. Black dots in dilatational quadrants are projected P-axes for each event and mechanisms labeled with an ‘‘s’’ are shallow events less than 10 km deep. Thick dashed lines show fault locations inferred from the tomography; thin dashed lines show locations or limits of faults inferred from Johnson et al. [1994, 1996, 1999]. Grey dashed boxes show areas used to select events for focal mechanism analysis. Abbreviations: DDMF—Darrington-Devils Mountain fault; SFZ— Seattle fault zone; SWIF—Southern Whidbey Island fault zone; TFZ—Tacoma fault zone. Stars show sampling locations from the 3-D model for construction of the Seattle 1-D velocity model; diamonds show sampling locations for construction of the North/East 1-D model (see text and Table 4). Pointers on axes show locations of cross-sections shown in Figure 9.

appear within the eastern portion of the Seattle basin, unlike [58] Focal mechanisms for 12 earthquakes with magni- the central and western portions, reflecting possible struc- tudes greater than 3.0 located outside the Tacoma, Seattle, tural and tectonic changes within the basin. and southern Whidbey Island fault zones were determined [57] The majority of focal mechanisms were determined (Figures 8 and 9), and all but one of these events appear to from earthquakes occurring in the Crescent basement. be located within the Cascade province. Mechanisms of Virtually no earthquakes occur in regions which we inter- these events are mainly oblique thrust and strike-slip with P- pret to be within the OSC, and relatively few earthquakes axes plunging shallowly and trending N–S. occur within the shallow sedimentary basins. Overall, focal [59] Two clusters of earthquakes in the northeastern part of mechanism solutions reveal primarily strike-slip and thrust the model are located in the vicinity of the eastern extent of mechanisms with P-axes trending N–S with an overall the Darrington-Devils Mountain fault (DDMF)—a region average plunge of less than 20° (Figure 10), suggesting that of very sparse seismicity. Three focal mechanisms were regional maximum compressive stress is aligned N–S and determined for the largest of these earthquakes (Figure 8; is nearly horizontal. These results are consistent with the Events 1, 11, and 14 of Table 3) which have depths between results of Ma et al. [1996]. 10 and 16 km. All three mechanisms are consistent with N–S ESE 22 - 18 VAN WAGONER ET AL.: CRUSTAL STRUCTURE AND EARTHQUAKES

Table 3. Relocated Hypocenters and Focal Solutions for Earthquakes Greater Than Magnitude 3.0 b b c Date, Latitude, Longitude, Depth, P-Axis T-Axis Plane A No.a UT °N °W km Magnitude Az Pl Az Pl Sk Dp Rk 1 22 August 1980 48.219 121.645 11.25 3.1 3 4 266 56 59 57 131 2 19 September 1980 47.908 121.851 5.22 3.8 135 4 225 5 0 89 6 3 21 September 1980 47.908 121.853 5.16 3.5 25 7 119 35 103 71 31 4 12 November 1981 47.952 122.405 26.73 3.7 107 11 345 69 32 58 111 5 12 April 1982 47.641 122.519 27.70 3.4 6 18 114 43 115 75 47 6 24 January 1983 47.083 121.967 4.02 3.0 353 18 255 21 56 88 152 7 25 May 1983 47.775 121.634 11.97 3.0 354 45 245 18 125 74 131 8 12 August 1983 47.504 121.644 15.16 3.1 297 17 107 73 155 62 87 9 5 October 1983 47.469 121.832 24.13 3.1 153 47 0 40 103 86 104 10 11 January 1985 47.911 122.887 21.31 3.3 178 15 36 72 97 61 102 11 5 March 1985 48.274 122.156 15.44 3.1 12 29 170 60 86 75 80 16 June 1985 47.439 121.840 17.83 3.1 6 July 1985 47.732 122.221 19.25 3.1 12 14 September 1985 47.393 122.387 18.68 3.0 327 2 236 12 79 83 170 13 30 December 1990 47.484 121.795 18.99 3.1 167 13 333 77 74 58 86 18 January 1992 47.405 122.710 13.18 3.1 14 20 January 1992 48.300 122.071 13.10 3.0 173 11 293 69 68 58 69 15 29 March 1992 47.630 122.196 27.83 3.1 193 12 67 70 116 59 108 16 26 January 1993 47.374 122.700 26.07 3.0 194 19 299 36 70 79 41 17 22 July 1994 47.634 122.091 8.83 3.2 12 46 192 44 102 89 90 18 10 September 1994 47.191 121.957 18.32 3.9 47 10 263 77 36 55 99 Robinson Point 19 29 January 1995 47.393 122.368 16.38 5.0 355 21 140 65 106 67 76 30 March 1996 47.322 122.433 14.77 3.0 Duvalld 20 3 May 1996 47.767 121.864 9.06 5.4 93 14 305 73 10 60 100 3 May 1996 47.776 121.878 9.17 3.0 21 10 February 1997 47.565 122.309 5.15 3.5 334 33 241 6 112 72 151 Bremerton 22 23 June 1997 47.602 122.572 11.45 4.9 31 49 235 38 134 85 102 23 27 June 1997 47.606 122.565 10.74 4.1 27 35 249 47 44 84 112 24 11 July 1997 47.597 122.548 10.08 3.4 24 16 268 55 43 67 123 25 12 February 1998 47.668 122.478 24.34 3.0 190 1 279 2 55 89 0 26 3 November 1998 47.525 122.751 22.77 3.1 17 27 260 41 44 82 127 27 4 January 1999 47.214 122.275 25.97 3.2 170 5 69 67 100 54 118 28 2 July 1999 47.374 122.401 21.47 3.0 353 20 120 59 115 69 65 29 1 November 2000 48.283 122.541 22.18 3.3 320 19 123 70 135 64 84 30 31 December 2000 47.515 121.646 14.98 3.0 359 14 138 70 102 60 75 a Label number refers to numbered focal mechanisms shown in Figures 8 and 9. bAxes orientation given as azimuth (Az) in degrees clockwise from the north and degrees plunge (Pl) from horizontal. cOrientation of nodal plane A from double-couple solution in degrees strike (Sk), dip (Dp), and rake (Rk) (plane A is not necessarily the preferred fault plane). Strike angles less than zero are measured counterclockwise from the north. dDuvall focal mechanism taken from PNSN catalog solution.

Table 4. 1-D Velocity Models Used to Compute Focal compression. The occurrence of these earthquakes suggests Mechanisms possible activity associated with the DDMF, which has Depth Range, Seattle Model North/East Model recently been postulated to be an active fault by Johnson et km P-Velocity, km/s P-Velocity, km/s al. [2001] on the basis of geologic and seismic reflection 0.0–0.5 3.20 3.62 data. However, the mechanisms of these deeper earthquakes 0.5–1.5 3.49 3.88 are not consistent with the left-lateral component of motion 1.5–2.5 3.78 4.13 postulated by Johnson et al. [2001] for the DDMF. Unfortu- 2.5–3.5 4.08 4.39 nately, our velocity model is poorly constrained in this region 3.5–4.5 4.37 4.64 4.5–5.5 4.66 4.90 due to proximity to the model boundary, preventing possible 5.5–6.5 4.96 5.16 delineation of the fault from velocity variation. 6.5–7.5 5.25 5.41 4.2.4. Tacoma Fault Zone (TFZ) 7.5–8.5 5.54 5.67 [60] On the western side of our model at depths less than 8.5–9.5 5.84 5.92 10 km (Figure 9a) a cluster of earthquakes appearing as a 9.5–10.5 6.13 6.18 10.5–11.5 6.42 6.44 subvertical lineation may be associated with the TFZ (coor- 11.5–14.5 6.46 6.48 dinate 60 N). If these events indeed reflect displacement on 14.5–17.5 6.49 6.53 the TFZ, they are consistent with a high angle, possibly 17.5–20.5 6.51 6.57 north-dipping, fault zone extending to at least 10 km depth 20.5–23.5 6.54 6.62 23.5–26.5 6.57 6.66 and possibly deeper. We suggest caution however, in inter- 26.5–29.5 6.59 6.71 preting such shallow nearly vertical distributions of hypo- 29.5–32.5 6.62 6.76 centers since depth constraint is often poor in the upper 10 32.5–35.5 6.65 6.80 km of the crust, and such vertical patterns may be, at least 35.5–40.0 6.67 6.85 VAN WAGONER ET AL.: CRUSTAL STRUCTURE AND EARTHQUAKES ESE 22 - 19

Figure 9. N–S cross-sections showing relocated earthquakes and computed focal mechanisms within ±12 km E–W of each respective section. Focal mechanisms are equal-area projections onto the half- sphere behind a N–S vertical plane and numbered mechanisms correspond to events listed in Table 3 and shown in Figure 8. Contours show velocity model (in km/s) at each vertical section. Section locations are shown in Figure 8 and are identical to those in Figure 5. Earthquake magnitude scale in lower left corner. Abbreviations: Br—Bremerton earthquake; KA—Kingston arch; MB—Muckleshoot basin; PB— Possession basin; PLVH—Port Ludlow velocity high; PTB—Port Townsend basin; RPt—Robinson Point earthquake; SB—Seattle basin; SFZ—Seattle fault zone; SqF—Sequim fault; SVH—Seattle velocity high; SWIF—Southern Whidbey Island fault zone; TB—Tacoma basin; TFZ—Tacoma fault zone.

partly, an artifact of location error. Unfortunately, only two predominantly thrust and strike-slip mechanisms, with focal mechanisms within the TFZ with hypocentral depths nodal planes striking E–W (Figures 8 and 9). P-axes for less than 10 km were well constrained; both are thrust the TFZ events are generally oriented N–S and dip 15° on solutions but are located approximately 10 km NE of the average; T-axes dip 55° on average and trend from west- probable location of the Tacoma fault (Figure 8). ward to vertical (Figure 10). The magnitude 5.0 Robinson [61] A total of 18 focal mechanisms were determined Point earthquake was relocated to a depth of 16.4 km. Our from earthquakes within the region of the TFZ, revealing mechanism for this event is virtually identical to the ESE 22 - 20 VAN WAGONER ET AL.: CRUSTAL STRUCTURE AND EARTHQUAKES

Figure 10. Equal-area, lower-hemisphere projections of P-axes (.) and T-axes (6) in the Seattle and Tacoma fault zone regions and all mechanisms done in this study. The Seattle and Tacoma fault zone regions are delineated by the appropriate dashed rectangles of Figure 8.

solution reported by Dewberry and Crosson [1996] (Event [65] Only three focal mechanisms within the SFZ are 19, Figures 8 and 9b and Table 3). from earthquakes shallower than 10 km (Figure 8). Two of 4.2.5. Seattle Fault Zone (SFZ) these shallow events have mechanisms similar to the Bre- [62] There is a greater frequency of shallow (<10 km merton earthquake—Event 21 is directly beneath the city of depth) seismic activity associated with the SFZ than with Seattle (Figures 8 and 9c and Table 3) and Event 17 is the TFZ. Seismicity is distributed over a broad zone and is slightly north of the deformation front defined by Blakely et not localized to one or more narrow faults, suggesting that al. [2002] with vertical and horizontal nodal planes with the SFZ is a complex structure. The western and eastern either north-side-up or thrust offset. The remaining shallow ends of the fault (Figures 9a and 9c) show increased activity event, at a depth of 5 km, is located beneath the east side of in the upper 10 km of the crust relative to the central portion Lake Washington and is almost purely strike-slip with nodal of the fault (Figure 9b). Seismicity in the SFZ does not planes oriented NNE and WNW (Figure 8). show evidence of alignment along the proposed south- [66] The most significant shallow earthquakes within the dipping thrust model of Johnson et al. [1994, 1999] and general region of the SFZ have occurred within or below Pratt et al. [1997]. the Seattle basin and are consistent with evidence of north- [63] A total of 26 focal mechanisms were determined for side-up displacement. The E–W trending Toe Jam Hill earthquakes near the SFZ; these are predominantly oblique fault scarp on southern Bainbridge Island [Blakely et al. thrust or strike-slip mechanisms (Figure 8). Nodal planes for 2002] is a north-dipping, north-side-up structure directly thrust mechanisms strike generally E–W with most solu- above the Bremerton earthquake sequence. Blakely et al. tions slightly rotated NW–SE. P-axes for the SFZ events are [2002] propose that the fault scarp is either the result of a generally oriented NNE or S and plunge shallowly at 18° on roof-thrust within the SFZ or is the result of movement on a average; T-axes have a broader plunge distribution averaging nearly vertical fault deep within the Seattle basin. We 42° and trending mostly west to vertical (Figure 10). believe that the latter interpretation is most consistent with [64] The magnitude 4.9 Bremerton earthquake was relo- the observed seismicity and focal mechanism evidence. cated in the joint inversion to a depth of 11.5 km. Our Both gravity evidence and corroborating seismic velocity computed focal mechanism for this event has one nearly evidence demonstrates that the Seattle basin is a large local vertical nodal plane striking NW with oblique normal offset mass deficiency, which is at least locally, and probably and one shallowly dipping nodal plane striking NNE with regionally, isostatically uncompensated. We suggest that oblique strike-slip offset (Event 22, Figures 8 and 9b and isostatic body force (buoyancy) may presently be elevating Table 3). Focal mechanisms for four aftershocks from the the basin producing displacement in or near the SFZ where Bremerton event reveal similarly oriented nodal planes as the basin is both deepest and fault bounded to a depth of at the main shock but with slightly larger strike-slip or thrust least 10 km. This scenario is consistent with the locations components (Figure 8). The location of the Bremerton and focal mechanisms of shallow events within the SFZ. earthquake relative to its aftershocks reveals a vertical Isostatic forces tending to elevate the Seattle basin must distribution of nearly 8 km with a lateral E–W scatter of coexist with arc-parallel tectonic compressive stress result- 3 km and virtually no lateral N–S variation. Lack of strong ing from oblique subduction as proposed by Wang et al. depth constraint for this sequence because of suboptimal [1997]. The relative roles of these two tectonic mechanisms station distribution may contribute in part to the vertical are unknown, but are clearly important in understanding the spread of hypocenters. Nevertheless, the aftershock pattern nature of the SFZ and the earthquake hazard that it is most consistent with the nearly vertical nodal plane as the presents. true fault. If the vertical plane ruptured, then the overall 4.2.6. Southern Whidbey Island Fault Zone (SWIF) motion is north-side-up and eastward with respect to the [67] Only one earthquake of magnitude greater than 3.0 south side of the fault. has occurred in the general vicinity of the SWIF since VAN WAGONER ET AL.: CRUSTAL STRUCTURE AND EARTHQUAKES ESE 22 - 21

1980 (Event 4, Table 3). We determined seven well earthquakes do not clearly delineate individual fault surfa- constrained mechanisms from a larger region containing ces, this localization of seismicity indicates a tectonically the SWIF (Figure 8). Hypocentral depths of these events active region of the Lowland which reinforces the potential are greater than 14 km and their focal mechanisms are seismic hazards associated with the Seattle and Tacoma mainly thrust with a variety of nodal plane orientations. fault zones. Although they may be related to the SWIF, we find no [73] Our investigation provides a reference 3-D P wave evidence in the location or orientations of these earth- velocity model (PS3DA) for future earthquake location quakes to directly correlate them to the fault, particularly studies, structure studies, and tectonic interpretation. because they occur significantly deeper than the region of Velocity variations within the Puget Lowland reveal a the fault zone imaged by seismic reflection profiles complex three-dimensional structural pattern that is not [Johnson et al., 1996]. easily explained by simple tectonic models. Computed earthquake focal mechanisms indicate that N–S regional tectonic compression is important and is active in both 5. Summary and Conclusions thrusting and strike-slip modes of deformation. However, [68] High-resolution 3-D seismic tomography using both many features such as apparent north striking faults along explosion and earthquake data provides a detailed P wave the east margin of the Port Ludlow velocity high and the velocity image of the Puget Lowland to a depth of 30 km. northern Tacoma basin, the concave-upward pattern of the The Lowland is characterized by six distinct basins varying basement beneath the Seattle basin, and the apparent in depth from 3 to 10 km (Figure 7). The basins are basin-side-up focal mechanisms within the Seattle fault separated by high velocity regions where Crescent forma- zone are not readily explained solely by N–S compres- tion rocks are typically within one or two kilometers of the sion, requiring the consideration of additional tectonic surface. In addition to the Seattle, Tacoma, Port Townsend, complexity. and Possession basins which have been previously described, we have described the Muckleshoot and Sequim [74] Acknowledgments. Research supported by the U.S. Geologi- cal Survey (USGS), Department of the Interior, under USGS award basins which have not formerly been delineated as distinct numbers 1434-HQ-97-GR-03084, 1434-HQ-98-00017 and 99-HQ-GR- basins. 0036, and the National Science Foundation under grant EAR-0107138 [69] The Seattle fault zone is a WNW trending zone of to Crosson and Creager. The views and conclusions contained in this rapid lateral velocity transition from the low velocity document are those of the authors and should not be interpreted as necessarily representing the official policies, either expressed or implied, sedimentary rocks of the Seattle basin to the high velocity of the U.S. Government. We thank the SHIPS Working Group (follow- region separating the Seattle and Tacoma basins. The ing individuals: M. A. Fisher, T. Parsons, R. A. Hyndman, K. C. Miller, combination of tomography, seismicity, and focal mecha- C. N. Snelson, D. C. Mosher, T. L. Pratt, R. Ramachandran, G. D. Spence, U. S. ten Brink, A. M. Trehu, C. S. Weaver, and B. C. Zelt) nisms suggest that the SFZ is a complex zone of deforma- for providing data, ensuring the success of the SHIPS experiments, and tion, rather than a single narrow fault, with north-side-up for advice and useful discussions. We would like to thank Harvey displacement perhaps driven by isostatic readjustment of the Greenberg for supplying digital geologic data, and Richard Stewart and Richard Blakely for helpful discussions. We are grateful to Ruth Ludwin low density basin. The SFZ is truncated on the west beneath and Brian Sherrod for early reviews of the paper, and for the comments Hood Canal and seismography provides no supporting of two anonymous reviewers who helped to significantly improve the evidence for extension of the fault eastward of Lake paper. Sammamish. [70] The Seattle and Port Ludlow velocity highs are strikingly analogous, suggesting similar processes of for- References mation. The Kingston arch appears to be an anticlinal fold Babcock, R. S., R. R. Burmester, K. P. Clark, D. C. Engebretson, and A. which at depth is the southeastward extension of the Port Warnock, A rifted margin origin for the Crescent basalts and related rocks Ludlow velocity high. Whether these velocity highs rep- in the northern Coast Range volcanic province, Washington and British Columbia, J. Geophys. Res., 97, 6799–6821, 1992. resent basement that has been structurally elevated, or Babcock, R. S., C. A. Suczek, and D. C. Engebretson, The Crescent ‘‘Ter- basement that has remained at relatively constant postem- rane’’, Olympic peninsula and southern Vancouver Island, Wash. Div. placement depth while the adjacent basins subsided cannot Geol. Earth Res. Bull., 80, 141–157, (Regional Geology of Washington), 1994. be directly ascertained by tomography but is a question Bakun, W. H., R. A. Haugerud, M. G. Hopper, and R. S. Ludwin, The remaining for further tectonic investigation. December 1872 Washington state earthquake (abstract), Seismol. Res. [71] Additional shallow faults inferred from our model Lett., 72, 269, 2001. include: the arcuate and apparently steeply dipping Tacoma Blakely, R. J., R. E. Wells, and C. S. Weaver, Puget Sound aeromagnetic maps and data, U.S. Geol. Surv. Open File Rep., 99-514, 1999. fault zone; N–S trending faults beneath Dabob Bay and to Blakely, R. J., R. E. Wells, C. S. Weaver, and S. Y. Johnson, Location, the east of the Port Ludlow velocity high; the arcuate structure, and seismicity of the Seattle fault zone: Evidence from aero- Sequim fault south of the Port Townsend basin; and a magnetic anomalies, geologic mapping, and seismic reflection data, Geol. Soc. Am. Bull., 114, 169–177, 2002. possible E–W trending fault north of the Sequim basin. Brandon, M. T., and A. R. Calderwood, High-pressure metamorphism and [72] Earthquakes jointly relocated during our inversion uplift of the Olympic subduction complex, Geology, 18, 1252–1255, are not significantly different on average from the 1-D 1990. Brandon, M. T., M. K. Roden-Tice, and J. I. Garver, Late Cenozoic ex- model locations. These earthquakes reveal a midcrustal humation of the Cascadia accretionary wedge in the Olympic mountains, zone of concentrated seismicity between 15 and 25 km northwest Washington state, Geol. Soc. Am. Bull., 110, 985–1009, depth occurring mainly within the Crescent basement which 1998. arches beneath the Lowland from its surface exposure on Brocher, T. M., and N. I. Christensen, Density and velocity relationships for digital sonic and density logs from coastal Washington and laboratory the Olympic Peninsula. Shallower hypocenters are found measurements of Olympic Peninsula mafic rocks and greywackes, U.S. near the Seattle and Tacoma fault zones. Although these Geol. Surv. Open File Rep., 01-264, 1–39, 2001. ESE 22 - 22 VAN WAGONER ET AL.: CRUSTAL STRUCTURE AND EARTHQUAKES

Brocher, T. M., and A. L. Ruebel, Compilation of 29 sonic and density logs Map, vol. 1, edited by D. B. Slemmons, E. R. Engdahl, M. D. from 23 oil test wells in western Washington state, U.S. Geol. Surv. Open Zoback, and D. D. Blackwell, pp. 77 –98, Geol. Soc. of Am., File Rep., 98-249, 60 pp., 1998. Boulder, Colo., 1991. Brocher, T. M., et al., Wide-angle seismic recordings from the 1998 Seismic Ma, L., R. S. Crosson, and R. S. Ludwin, Western Washington earthquake Hazards Investigation of Puget Sound (SHIPS), western Washington and focal mechanisms and their relationship to regional tectonic stress, in British Columbia, U.S. Geol. Surv. Open File Rep., 99-314, 110 pp., Assessing Earthquake Hazards and Reducing Risk in the Pacific North- 1999. west, vol. 1, U.S. Geol. Surv. Prof. Pap., vol. 1560, edited by A. M. Brocher, T. M., et al., Report for explosion and earthquake data ac- Rogers, T. J. Walsh, W. J. Kockelman, and G. R. Priest, pp. 257–283, quired in the 1999 Seismic Hazards Investigation of Puget Sound 1996. (SHIPS), Washington, U.S. Geol. Surv. Open File Rep., 00-318,85 Malone, S., R. S. Crosson, K. C. Creager, A. Qamar, G. C. Thomas, R. pp., 2000. Ludwin, K. G. Troost, D. B. Booth, and R. A. Haugerud, Preliminary Brocher, T. M., T. Parsons, R. J. Blakely, N. I. Christensen, M. A. Fisher, report on the MW = 6.8 Nisqually, Washington earthquake of 28 February and R. A. Wells, Upper crustal structure in Puget Lowland, Washington: 2001, Seismol. Res. Lett., 72, 353–362, 2001. Results from the 1998 Seismic Hazards Investigation in Puget Sound, Malone, S. D., and S. S. Bor, Attenuation patterns in the Pacific Northwest J. Geophys. Res., 106, 13,541–13,564, 2001. based on intensity data and the location of the 1872 North Cascades Bucknam, R. C., E. Hemphill-Haley, and E. B. Leopold, Abrupt uplift earthquake, Bull. Seismol. Soc. Am., 69, 531–546, 1979. within the past 1700 years at southern Puget Sound, Washington, Science, Massey, N. W. D., Metchosin igneous complex, southern Vancouver Island: 258, 1611–1614, 1992. Ophiolite stratigraphy developed in an emergent island setting, Geology, Bucknam, R. C., B. L. Sherrod, and G. Elendahl, A fault scarp of probable 14, 602–605, 1986. Holocene age in the Seattle fault zone, Bainbridge Island, Washington Miller, K. C., G. R. Keller, J. M. Gridley, J. H. Luetgert, W. D. Mooney, and (abstract), Seismol. Res. Lett., 70, 233, 1999. H. Thybo, Crustal structure along the west flank of the Cascades, western Christensen, N. I., and W. D. Mooney, Seismic velocity structure and Washington, J. Geophys. Res., 102, 17,857–17,873, 1997. composition of the continental crust: A global view, J. Geophys. Res., Molzer, P. C., U. S. ten Brink, M. A. Fisher, T. M. Brocher, K. C. Creager, 100, 9761–9788, 1995. and R. S. Crosson, Seismic structure of the Seattle fault, Seattle basin, Clowes, R. M., M. T. Brandon, A. G. Green, C. J. Yorath, A. S. Brown, and Kingston arch, Washington (abstract), Eos Trans. AGU, 80(46), E. R. Kanasewich, and C. Spencer, LITHOPROBE-Southern Vancouver F762, Fall Meet. Suppl., 1999. Island: Cenozoic subduction complex imaged by deep seismic-reflec- Nelson, A. R., S. K. Pezzopane, and R. C. Bucknam, Holocene surface tions, Can. J. Earth Sci., 24, 31–51, 1987. faulting in the Seattle fault zone, Bainbridge Island, Washington (ab- Crosson, R., Crustal structure modeling of earthquake data, 2, Velocity stract), Seismol. Res. Lett., 70, 233, 1999. structure of the Puget Sound region, Washington, J. Geophys. Res., 81, Parsons, T., R. E. Wells, and M. A. Fisher, Three-dimensional velocity 3047–3054, 1976. structure of Siletzia and other accreted terranes in the Cascadia forearc Danes,Z.F.,M.M.Bonno,E.Brau,W.D.Gilham,T.F.Hoffman, of Washington, J. Geophys. Res., 104, 18,015–18,039, 1999. D. Johansen, M. H. Jones, B. Malfait, J. Masten, and G. O. Teague, Pratt, T. L., S. Johnson, C. Potter, W. Stephenson, and C. Finn, Seismic Geophysical investigation of the southern Puget Sound area, Washington, reflection images beneath Puget Sound, western Washington state: The J. Geophys. Res., 70, 5573–5580, 1965. Puget Lowland thrust sheet hypothesis, J. Geophys. Res., 102, 27,469– Dewberry, S. R., and R. S. Crosson, The MD 5.0 earthquake of 29 27,489, 1997. January 1995 in the Puget Lowland of western Washington: An Ramachandran, K., Velocity structure of S.W. British Columbia, and event on the Seattle fault?, Bull. Seismol. Soc. Am., 86, 1167–1172, N.W. Washington, from 3-D non-linear seismic tomography, Ph.D. 1996. thesis, 198 pp., Univ. of Victoria, Victoria, British Columbia, Canada, Finn, C., W. M. Phillips, and D. L. Williams, Gravity anomaly and terrain 2001. maps of Washington, U.S. Geol. Surv. Geophy. Invest. Ser. Map, GP-988, Schruben, R. A., R. E. Arndt, W. J. Bawiec, and R. A. Ambroziak, Geology scale 1:1,000,000 and 1:500,000, 1991. of the coterminous United States at 1:2,500,000 scale: A digital repre- Fisher, M. A., et al., Results from the 1998 SHIPS Experiment, regarding sentation of the 1974 P. B. King and H. M. Beikman map, U.S. Geol. urban earthquake hazards in Cascadia (abstract), Eos Trans. AGU, Surv., Denver, Colo., 1994. 80(46), F762, Fall Meet. Suppl., 1999. Schultz, A. P., and R. S. Crosson, Seismic velocity structure across the Gower, H. D., J. C. Yount, and R. S. Crosson, Seismotectonic map of the central Washington Cascade range from refraction interpretation with Puget Sound region, Washington, U.S. Geol. Surv. Misc. Invest. Ser. Map, earthquake sources, J. Geophys. Res., 101, 27,899–27,915, 1996. I-1613, scale 1:250,000, 1985. Sherrod, B. L., Late Holocene environments and earthquakes in southern Haeussler, P. J., and K. P. Clark, Geologic map of the Wildcat Lake 7.50 Puget Sound, Ph.D. thesis, 159 pp., Univ. of Wash., 1998. quadrangle, Kitsap and Mason counties Washington, U.S. Geol. Surv. Snavely, P. D., and E. M. Baldwin, Siletz River volcanic series, northwestern Open File Rep., 00-356, scale 1:24,000, 1 sheet, 2000. Oregon, Bull. Am. Assoc. Petrol. Geol., 32, 805–812, 1948. Johnson, S. Y., Evidence for a margin-truncating transcurrent fault Snavely, P. D., Jr., and R. E. Wells, Cenozoic evolution of the continental (pre-late Eocene) in western Washington, Geology, 12,538–541, margin of Oregon and Washington, in Assessing Earthquake Hazards 1984. and Reducing Risk in the Pacific Northwest, vol. 1, edited by A. M. Johnson, S. Y., Eocene strike-slip faulting and nonmarine basin for- Rogers, T. J. Walsh, W. J. Kockelman, and G. R. Priest, U.S. Geol. Surv. mation in Washington, in Strike-Slip Deformation, Basin Formation, Prof. Pap., vol. 1560, pp. 161–182, 1996. and Sedimentation, no. 37, edited by K. T. Biddle and N. Chris- Symons, N. P., Seismic velocity structure of the Puget Sound region from tie-Blick, Spec. Pub. Soc. Econ. Paleon. Mineral., pp. 283–302, 3-D non-linear tomography, Ph.D. thesis, 168 pp., Univ. of Washington, 1985. Seattle, Wash., 1998. Johnson, S. Y., C. J. Potter, and J. M. Armentrout, Origin and evolution of Symons, N. P., and R. S. Crosson, Seismic velocity structure of the Puget the Seattle Fault and Seattle Basin, Washington, Geology, 22, 71–74, Sound region from 3-D non-linear tomography, Geophys. Res. Lett., 24, 1994. 2593–2596, 1997. Johnson, S. Y., C. J. Potter, J. M. Armentrout, J. J. Miller, C. Finn, and C. Tabor, R. W., Late Mesozoic and possible early Tertiary accretion in wes- S. Weaver, The southern Whidbey Island fault: An active structure in tern Washington state—the Helena-Haystack melange and the Darring- the Puget Lowland, Washington, Geol Soc. Am. Bull., 108, 334–354, ton-Devils Mountain fault zone, Geol. Soc. Am. Bull., 106, 217–232, 1996. 1994. Johnson, S. Y., S. V. Dadisman, J. R. Childs, and W. D. Stanley, Active Tabor, R. W., and W. M. Cady, Geologic map of the Olympic Peninsula, tectonics of the Seattle Fault and central Puget Sound, Washington-Im- Washington, U.S. Geol. Surv. Misc. Invest. Ser., Map I-994, scale plications for earthquake hazards, Geol Soc. Am. Bull., 111, 1042–1053, 1:125,000, 1978a. 1999. Tabor, R. W., and W. M. Cady, The structure of the Olympic Mountains, Johnson, S. Y., S. V. Dadisman, D. C. Mosher, R. J. Blakely, and J. R. Washington-analysis of a subduction zone, U.S. Geol. Surv. Prof. Pap., Chiles, Active tectonics of the Devils Mountain fault and related struc- 1033, 38 pp., 1978b. tures, northern Puget lowland and eastern Strait of Juan de Fuca region, Tabor, R. W., V. A. Frizzell Jr., D. B. Booth, R. B. Waitt, J. T. Whetten, and Pacific Northwest, U.S. Geol. Surv. Prof. Pap., 1643, 45 pp., 2 sheets, R. E. Zartman, Geologic map of the Skykomish River 30- by 60-minute 2001. quadrangle, Washington, U.S. Geol. Surv. Misc. Invest. Ser., Map I-1963, Lees, J. M., and R. S. Crosson, Tomographic imaging of local earthquake scale 1:100,000, 1993. delays times for three-dimensional velocity variation in western Washing- ten Brink, U. S., P. C. Molzer, M. A. Fisher, R. J. Blakely, R. S. Crosson, ton, J. Geophys. Res., 95, 4763–4776, 1990. and K. C. Creager, Subsurface geometry and evolution of the Seattle Ludwin, R. S., C. S. Weaver, and R. S. Crosson, Seismicity of Wa- fault zone and the Seattle basin, Bull. Seismol. Soc. Am., 92, 1737– shington and Oregon, in Neotectonics of North America Decade 1753, 2002. VAN WAGONER ET AL.: CRUSTAL STRUCTURE AND EARTHQUAKES ESE 22 - 23

Trehu, A. M., I. Asudeh, T. M. Brocher, J. H. Luetgert, W. D. Mooney, J. L. Yount, J. C., and H. D. Gower, Bedrock geologic map of the Seattle 300 by Nabelek, and Y. Nakamura, Crustal architecture of the Cascadia forearc, 600 quadrangle, Washington, U.S. Geol. Surv. Open File Rep., 91-147,37 Science, 266, 237–243, 1994. pp., scale 1:100,000, 5 sheets, 1991. Wang, K., Simplified analysis of horizontal stresses in a buttressed forearc sliver at an oblique subduction zone, Geophys. Res. Lett., 23, 2021– 2024, 1996. T. M. Brocher, U.S. Geological Survey, 345 Middlefield Road, MS-977, Wang, K., J. He, and E. E. Davis, Transform push, oblique subduction Menlo Park, CA 94025, USA. resistance, and intraplate stress of the Juan de Fuca plate, J. Geophys. K. C. Creager, R. S. Crosson, G. Medema, and L. Preston, Department of Res., 102, 661–674, 1997. Earth and Space Sciences, University of Washington, Box 351310, Seattle, Wells, R. E., D. C. Engebretson, P. D. Snavely Jr., and R. S. Coe, Cenozoic WA 98195-1310, USA. ([email protected]) plate motions and the volcano-tectonic evolution of western Oregon and N. P. Symons, Sandia National Laboratories, Box 5800, MS-0750, Washington, Tectonics, 3, 275–294, 1984. Albuquerque, NM 87185-0750, USA. Wells, R. E., C. S. Weaver, and R. J. Blakely, Fore-arc migration in Cascadia T. M. Van Wagoner, ChevronTexaco North America Upstream, 935 and its neotectonic significance, Geology, 26, 759–762, 1998. Gravier Street, New Orleans, LA 70112, USA. ([email protected])